You don't own your software — you rent it, and the rent compounds: the SaaS subscription tax

The subscription model won because the first invoice is small. Instead of paying to build or buy a system outright, you pay a little per user per month and you are live by Friday — no big cheque, no project, no risk. That trade is genuinely good when you are starting out, and it is why almost every business now runs on rented software. The catch is hidden in the shape of the deal. A subscription is not a price; it is a curve. It starts low and it bends upward, because the thing you are renting is priced per seat and per month — so it grows with your headcount, with the number of tools you add to cover the gaps, and with every renewal that quietly steps up. You did not buy a tool. You signed up to pay rent for as long as you are in business, on terms that get more expensive precisely as you succeed.

At the level of the whole business, that curve has gotten steep. The average company now runs 106 separate SaaS applications [1], and worldwide spending on them passed 299 billion dollars in 2025, up almost 20 percent in a single year [2]. The number that should stop an operator cold is not the total, though — it is the leak. Across organisations, roughly 36 percent of the SaaS licenses they pay for are never used, which works out to about 19.8 million dollars a year, per organisation, spent on software nobody opens [3]. Gartner puts the same finding in plainer terms: about a quarter of every dollar spent on SaaS goes to unused or underused software [4]. You are not just renting; you are renting more than you use, and you cannot see how much because the bills arrive separately, monthly, from a dozen different vendors.

This piece is not an argument against SaaS — renting is the right call for plenty of things, and we will say where. It is an argument about ownership: about understanding that a subscription stack is a cost curve you have agreed to ride upward forever, that a meaningful slice of it is pure waste by the time you are a real company, and that for the systems your business actually runs on, there is a different curve available — one you own, that bends the other way.

Start with why the rent compounds, because the per-seat model is engineered to. When you rent software, the vendor's price is tied to the two things you most want to grow: how many people use it and how long you keep using it. So the bill rises automatically with your headcount — every hire is another seat on every tool they touch — and it rises again at renewal, when last year's introductory rate steps up. Then it rises a third way that nobody plans: tool sprawl. A subscription is cheap enough that any team can add one to fix a single problem, so they do, and the average company ends up running 106 of them ^[1]. Each one is individually defensible and collectively a fortune, and because they are billed separately, no single invoice ever looks alarming. That is the trap in one sentence: the cost is large, but it never arrives large.

Now the part that turns a fair deal into a bad one — the waste. Independent measurement of real SaaS portfolios finds that roughly 36 percent of the licenses companies pay for are simply never used: the employee left, the project ended, the team switched tools, but the seats keep billing ^[3]. That is not an edge case; it averages out to about 19.8 million dollars a year for a large organisation, and about 1,739 dollars per employee per year, spent on software nobody opens ^[3]. Gartner reaches the same place from the top down — roughly 25 percent of all SaaS spend goes to unused or underused entitlements and overlapping tools ^[4]. Read those two numbers together and the picture is stark: you are riding a cost curve upward, and a quarter to a third of what you pay at every point on that curve is buying you nothing at all.

The reason the waste survives is structural, not careless. When your business runs on rented tools, no one owns the total. Finance sees a row of small recurring charges; each department sees a tool it needs; nobody sees the 106-app portfolio as one number with a third of it dark. Worse, you usually cannot just turn the waste off — multi-year contracts, auto-renewals, and the data trapped inside each tool mean cancelling an app you have outgrown is itself a project ^[1]. So the leak persists not because people are wasteful but because the architecture of renting hides the total and raises the cost of leaving. The subscription that was frictionless to start becomes expensive to stop.

Set against that is the curve almost nobody draws on the whiteboard: ownership. When you build the system your business actually runs on, the cost arrives the other way around — high at the start, because building is real work, then almost flat, because the marginal cost of the next user, the next record, the next small feature is close to zero. You are not paying per seat; you are paying for infrastructure that does not care whether ten people or a hundred use it. The two curves — rising rent, flat ownership — cross somewhere, and where they cross is the whole decision. For a tool you will use lightly for a year, the rent never reaches the crossover and renting wins easily. For the system at the centre of your business that you will run for a decade and grow headcount on, the rented curve sails past the build cost and keeps climbing, while the owned one sits flat underneath it ^[2].

So the operator's test is not "build or buy" in the abstract — it is which side of the crossover each system lives on. Three questions sort it fast. First, is this core or peripheral? Rent the peripheral things — email, video calls, accounting, the tools every business uses the same way; there is no edge in owning them and the vendor's scale beats yours. Own the things that are how your business specifically works: the CRM that holds your pipeline, the operations system that runs your jobs, the customer-facing app that is your product. Second, how long and how wide? The longer you will run it and the more seats it will grow to, the further past the crossover renting carries you. Third, what does leaving cost? If your most important data lives inside someone else's tool on someone else's terms, you do not own your business's memory — you are renting that too. The subscription model is a brilliant way to start and a quietly punishing way to run a business at scale. Rent what is common and cheap to leave. Own what is core, long-lived, and yours — and stop paying rent on the curve that only bends one way.